Aider Review 2026: Free Terminal AI Coding, Models & Limits

Disclosure: CodeToolPilot may earn a commission if you buy through links on this page, at no extra cost to you. How we make money.

Last verified: October 13, 2026. This is an editorial research review — how we review. Prices and plans change often; confirm on the vendor’s site.

Aider is an open-source AI pair programmer that runs in your terminal. You start it inside a git repository, add the files you want to work on and describe the change in plain English. Aider sends the request to the large language model you have chosen, applies the edits to your files and commits them to git with a descriptive message. It is released under the Apache 2.0 license and developed in the open under the Aider-AI organization on GitHub, where it has close to 50,000 stars.

Aider is for developers who are comfortable on the command line, want to pick their own models and pay their provider directly, and like seeing every AI edit as an ordinary git commit they can diff or undo. There is no subscription, no account and no vendor-hosted service. The software is free, and your only costs are what your LLM provider charges, or your own hardware if you run local models.

This is an editorial research review based on Aider’s documentation, release history, GitHub repository and PyPI listing. We did not run our own benchmarks or timed tests.

Quick verdict

Aider is still one of the most transparent and least locked-in ways to point an LLM at a real codebase. Its git-first workflow, repository map and broad model support are well documented, and the tool itself costs nothing. The trade-offs are a noticeably slower pace of development since mid-2025, a terminal-centric workflow, and none of the managed cloud agents, team administration or enterprise controls that commercial coding agents now bundle.

Pro: Free and open source under Apache 2.0. You pay only your model provider, or nothing beyond hardware with local models.
Pro: Works with most LLMs, including OpenAI, Anthropic, Gemini, DeepSeek, xAI, Amazon Bedrock, Vertex AI, OpenRouter, Ollama, LM Studio and any OpenAI-compatible API.
Pro: Git-native. Every edit is committed with a descriptive message, and /diff and /undo keep changes easy to review and reverse.
Pro: A repository map of files, classes and function signatures gives the model context across larger codebases.
Con: Development has slowed. The latest PyPI release, 0.86.2, came out in February 2026, and the last feature release (0.86.0) in August 2025.
Con: Terminal-first. Editor use works through file-watching comments rather than a native IDE panel.
Con: No team plan, admin console, SSO or vendor support contract.
Con: Results and costs depend heavily on the model you choose, and the docs warn that weaker models may struggle to make edits.

Visit Aider

What is Aider?

Aider is a Python command-line application that edits files in place and commits each change. It can start a new project or work inside an existing codebase. Its homepage says it supports more than 100 programming languages, and installers are documented for macOS, Linux and Windows.

Aider also publishes LLM leaderboards based on a polyglot benchmark of 225 Exercism exercises in six languages. They were last updated on November 20, 2025, so they do not cover the newest models. According to the FAQ, Aider usually writes about 70% of the new code in each of its own releases.

Aider is live and has not been discontinued, but its release pace has slowed. According to PyPI, version 0.86.0 shipped on August 9, 2025, and the patch release 0.86.2 on February 12, 2026. The GitHub main branch was still receiving commits into May 2026, including model settings for newer GPT and Claude models. Those changes had not reached a PyPI release when we checked.

Key features

Git integration and undo

Aider works best in a git repository and offers to create one if none exists. It commits each edit with a generated message, in Conventional Commits style by default, and commits any existing uncommitted changes first so your work stays separate. /diff shows recent changes and /undo reverts the last one. Commits are marked “(aider)”, and auto-commits can be turned off.

Repository map

With each request, Aider sends the model a compact map of the repository: the files and their most important classes and functions, with signatures. A graph-ranking step chooses the most relevant parts to fit a token budget, which defaults to 1,000 tokens and can be changed with --map-tokens. The model can use existing APIs without opening every file, and it can ask for more files when it needs them.

Chat modes, including architect mode

Code mode, the default, edits files. Ask mode discusses the code without changing anything, and help mode answers questions about Aider itself. Architect mode splits each task between two models: a main model proposes the solution and a separate editor model turns it into precise file edits. The docs note that this means two requests per task, so it can be slower and cost more.

Bring any model

Aider’s docs list connections for OpenAI, Anthropic, Gemini, Groq, xAI, Azure, Cohere, DeepSeek, OpenRouter, Vertex AI, Amazon Bedrock, Ollama and LM Studio, plus any OpenAI-compatible API. It can also call the GitHub Copilot API with an existing Copilot subscription.

Lint, test and fix

Aider lints the files it edits by default, using built-in linters for popular languages or your own --lint-cmd. With --test-cmd and --auto-test, it runs your test suite after each change and tries to fix any failures.

IDE comments, images, web pages and voice

With --watch-files, Aider watches the repository for comments ending in AI! (make a change) or AI? (answer a question), so you can trigger it from any editor. You can add images to the chat for vision-capable models, scrape a web page into context with /web, and dictate requests with /voice.

Who should use Aider

Aider suits individual developers and small teams who live in the terminal, want full control over which model sees their code, and value a clean git history. It is a good fit if you already have API keys or a Copilot subscription, want to experiment with local models, or need a free, open-source tool you can inspect and script.

Who should skip it

Skip Aider if you want a polished IDE experience, an agent that runs autonomously in the cloud, or a vendor-backed product with SSO, admin controls and a support contract.

Strengths and limitations

Strengths

No lock-in. You can switch models or providers with a flag, run fully local, and leave at any time with nothing but ordinary git history behind you.
Reviewable changes. Committing every edit keeps AI changes separate, attributable and easy to inspect or roll back with standard git tools.
Cost transparency. With no markup or subscription, you pay provider list prices, and Aider prints estimated costs as you work.
Privacy-minded defaults. Analytics are opt-in, and the docs say Aider never collects your code, chat messages, keys or personal information. Your code goes only to the model provider you configure.

Limitations

Slower development. No feature release since August 2025 and only a patch release in 2026. Newer model settings sit on the main branch, and the leaderboards stop at November 2025. That matters if you need quick support for new models.
Dated model advice. The homepage and model docs still recommend older models, so you need to research current options yourself.
Large repositories. The FAQ says Aider “is not optimized for quick performance and response time in very large repos” and suggests --subtree-only and .aiderignore to narrow the scope.
Less autonomy than newer agents. Aider works best when you choose which files go into the chat. It does not offer cloud agents, parallel background tasks or pull request automation.
No organizational features. There is no central policy, usage dashboard, SSO or support agreement. The Python scripting API is described as unsupported and subject to change.

Pricing (as of October 2026)

Plan	Price	What you get
Aider (open source)	$0	The full tool under the Apache 2.0 license. No account, subscription or paid tier.
Cloud model APIs	Billed by your provider	You bring API keys for OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter or another provider and pay its token rates. OpenRouter offers free access to some models with daily limits.
Local models	No API cost	Run models through Ollama, LM Studio or any OpenAI-compatible local server. You supply the hardware.
GitHub Copilot subscription	Your existing Copilot plan	Aider can call Copilot’s API, with usage billed to your Copilot subscription.

Aider has no pricing page. Your spending depends on the model, your provider’s rates and how much context you send. Adding extra files or using architect mode increases token use. Confirm current rates on your model provider’s pricing page. See Aider’s documentation

Prices verified on October 13, 2026 against the official Aider pricing page. Plans change often — confirm on the vendor’s page before you buy.

Use cases

Aider works best when a developer stays in the loop, chooses the files and reviews each commit. These scenarios are based on Aider’s documentation and the research behind our review, not on hands-on testing.

Focused changes in an existing codebase with a clean git trail

A backend developer needs to add an endpoint, update its tests and refactor a helper across a few files. Aider edits the files they add to the chat, uses its repository map to see related code, and commits each change with a descriptive message, so the work can be reviewed with git diff or rolled back with /undo. The caveat is that results depend on the model and on choosing the right files, and the docs warn that adding unnecessary files can confuse the model and raise costs.

AI help on code that should stay local

A developer working on a private codebase wants AI help without sending code to a cloud provider. Aider can connect to local models through Ollama, LM Studio or any OpenAI-compatible server, and its analytics are opt-in and, per its docs, never include code or chat messages. In our assessment, this is a reasonable fit for privacy-sensitive work, though local models are often less capable at editing, and the docs warn that weaker models may struggle.

Test-driven fix loops

A developer fixing a failing build wants the AI to keep going until the tests pass. With --test-cmd and --auto-test, Aider runs the test suite after each edit and tries to fix the failures, and it lints edited files by default. This works for compiled languages too, by pointing the test command at a build. Each iteration costs tokens, so it is worth watching spend on a paid API.

Batch edits from a script

A maintainer wants to apply the same mechanical change, such as adding docstrings, across many files. Aider’s --message option runs a single instruction against named files and exits, so it can be looped over in a shell script. Review the resulting commits before merging, and note that the Python scripting API is unofficial and may change.

Alternatives

Claude Code: Anthropic’s agent for the terminal, IDE, desktop and cloud. It is more autonomous and ships frequently, but it is tied to Claude models and Anthropic plans.
OpenAI Codex: an open-source CLI, IDE extension and cloud agent that comes with ChatGPT subscriptions.
Cline: an open-source, bring-your-own-model agent for VS Code, JetBrains, the CLI and desktop. It is the closest match if you want Aider’s model freedom inside an editor.
Amp: a multi-model coding agent with parallel cloud agents, for teams that want more automation.

Final verdict

Aider is a free, open, git-first terminal pair programmer that works with almost any model and keeps every AI change reviewable. Its design is still sound, but its development has slowed noticeably, so it is best treated as a mature tool rather than one at the frontier. If you want low cost and full control and don’t mind the command line, it is well worth trying. If you need managed agents or enterprise controls, look at the alternatives above.

Get started with Aider

FAQ

Is Aider free?

Yes. Aider is open source under the Apache 2.0 license, with no paid tier or account. You pay whichever LLM provider you connect through your own API keys, or nothing beyond hardware if you run local models. It can also use an existing GitHub Copilot subscription, with usage billed to that plan.

Which models work with Aider?

Aider’s docs say it can connect to most LLMs, including OpenAI, Anthropic, Gemini, DeepSeek, xAI, Groq, Azure, Amazon Bedrock, Vertex AI, OpenRouter, Ollama, LM Studio and OpenAI-compatible APIs. Some of the “works best with” recommendations in the docs list older models, and the official leaderboards were last updated in November 2025, so check current model choices yourself.

Is Aider still maintained?

Yes, but more slowly than before. PyPI shows the last feature release, 0.86.0, in August 2025 and a patch release, 0.86.2, on February 12, 2026, and the GitHub main branch received commits into May 2026. There is no discontinuation notice, but teams that need quick support for new models should take the pace into account.

Does Aider send my code anywhere?

Aider sends the files in the chat, plus a repository map, to the model provider you configure. Its own analytics are opt-in, and the docs say they never collect your code, chat messages, keys or personal information. To keep code entirely on your machine, use a local model.

Can I use Aider inside VS Code or JetBrains?

There is no official IDE extension. Instead, run Aider in a terminal with --watch-files and write comments ending in AI! or AI? in any editor to trigger edits or questions. If you want an agent built into the editor, see our Cline review.

Aider vs Claude Code: which should I choose?

Both are terminal coding agents. Claude Code is a commercial product tied to Claude models and Anthropic plans, with more autonomy, IDE and cloud options, and admin controls. In our assessment, Aider is the better fit if you want a free, open-source tool that works with many model providers and commits every edit to git, and Claude Code suits teams that want a faster-moving, vendor-supported agent.

Aider Review (2026): Open-Source AI Pair Programming in Your Terminal